Custodian vs RTA: Two Registers at Opposite Ends

One party keeps a scheme's securities and a record of them. Another keeps the list of folios and unit balances. The first shows what sits inside the pool; the second shows whose claim the pool answers. Nothing in either produces the other, and a single number, the value per unit, is where they meet.

The shape is far older than any scheme, and an everyday version of it will do to start. A caterer stores the crockery belonging to a residents association in a locked room and keeps a written list of every plate in it. A separate clerk keeps the ledger of which of the two hundred flats paid the levy and what share each one is due. Asked how many plates the flat on the third floor is entitled to, the storekeeper cannot say, however carefully he has counted. Asked what state the plates are in, the clerk cannot say either. Neither of them is being unhelpful. Each is looking at a different end of one arrangement, and it was set up that way deliberately.

One scheme runs through everything below. Girnar Asset Management Limited, an invented asset manager, runs the Girnar Large Cap Equity Fund. The scheme is open ended, invests in equity, and sits on Rs 4,200 crore of net assetsThe whole of what a scheme holds, less the whole of what it owes, on the day the number is struck. against 120.00 crore units in issue and spread over 3,80,000 foliosThe numbered account against which one holder's units in a scheme are written down.. Kalyani Bhagat has the portfolio; the operations desk answers to Sohail Merchant. Each service party is named by the role it fills rather than by a firm, and either role can be filled by any number of firms without the job itself changing at all.

Three things are set out separately. The operations map for a scheme fixes who is able to change what. The registrar and transfer agent, covered separately, keeps the register itself. Asking for a change and making one are two separate powers, and readers find that split the hardest: no distributor, adviser or platform can rewrite a row, move units or alter a count. Reconciliation, also covered separately, sets out the matches a scheme runs between its records, and why the two sides of a match must arrive from two places. Stand the two record keepers next to each other, and the reason for splitting them falls straight out.

Which of the two keeps securities, and which keeps only a list?

Most of the muddle on this subject comes from one word being made to do two jobs, so take the two parties one at a time and be exact. The custodianThe party in whose charge a scheme leaves its securities, kept clear of anything belonging to the manager. has charge of the securities the scheme has bought. Not a description of them, not an undertaking about them: the things themselves, kept clear of anything belonging to the asset manager, alongside the custody recordThe itemised list of what a custodian has in its charge, one line to a holding. saying which ones and how many. The registrar and transfer agentThe party looking after folio records and unit balances, and working through what holders send in. has charge of no securities whatever. The one thing it keeps is the unit registerThe roll of folios in a scheme, each line carrying the units standing to that folio., a roll of folios with a unit balance written against each.

So what separates them is not size, seniority or importance, and reaching for any of those three is exactly what sends a reader wrong. One party keeps a thing and the record of that thing; the other keeps the record and nothing else, and that lone asymmetry runs through everything that follows. A cloakroom at a wedding hall is the same shape in miniature. The counter has the coats and the counterfoil book saying which peg each one hangs on. The receptionist at the gate has the guest list and nothing more. Between them they cover a great deal; neither can supply what the other knows.

Neither party can do the other's job, and the reason is that neither one is given the other's input. The registrar and transfer agent has nothing to count and no valuation method to put to work on anything. The custodian has no folio lines in front of it to add. Either would first have to be handed the record the other keeps before it could produce the other's answer, and at that moment both records would sit with one keeper and the split would be gone. The arrangement can look like a share out of labour that a capable enough party could simply swallow. The share out is not of labour but of inputs.

Why do the two records describe one pool from opposite ends?

Because each was built to settle a question the other was never put. The custody record of the Girnar Large Cap Equity Fund yields a statement about contents: this many of security P, that many of security Q, this much cash lodged at a bank, this much due in and this much due out. The unit register of the same scheme yields a statement about claims: folio A stands at so many units, folio B at some other number, and 3,80,000 lines cut to that same pattern. One pool, two descriptions, and at no point do the two descriptions touch.

Neither record can be worked out from the other, in either direction, and piling on detail does not move that an inch. Set down the custody record security by security, to the last fraction of the smallest line, and there is still no folio number in it anywhere. Asked what balance stands on folio A, the custody record offers nothing to work from. The input the question needs, the folio lines, is not in it. The other direction fails the same way. No security appears in the register either, so all 3,80,000 register lines together still do not say how many of security P the scheme has bought.

Which settles a fair question: why does a scheme want two record keepers instead of one bigger and better resourced version of either? A larger custody record is a longer list of contents. A larger register is a longer list of claims. However far either grows it grows away from the other, so the second record is neither duplication nor a backup. It is the sole source of an answer the first one has no structural way to produce.

What single quantity joins the two records?

Exactly one, and it deserves naming before it is worked. The join is made by the value per unitWhat one unit comes to once the pool is shared out over the units standing in issue, and the number a scheme publishes., together with the relation that puts it there. Net assets sit at the custody end. Units in issue sit at the register end. Divide the one by the other and out comes the number both ends are describing at the same moment from different sides. The value per unit is the only spot at which the two records make contact. Any check reaching across them has to travel through it.

Work it on the scheme. Rs 42,00,00,00,000/- of net assets is the Rs 4,200 crore figure written out in full rupees, and 1,20,00,00,000 units in issue is 120.00 crore written the same way. The first divided by the second gives Rs 35.00 a unit. Not approximately Rs 35.00 and not rounded to it: the division comes out whole, with nothing left over. Now go back the other way. 1,20,00,00,000 units at Rs 35.00 a unit is Rs 42,00,00,00,000/-, the figure the first division started from.

Here is the part that is easy to get wrong, and it matters more than it looks. Doing the multiplication back feels like a check, and it is not one. The two directions are one equation rearranged, not two independent tests. Write N for net assets, U for units in issue and V for the value per unit. V is defined as N over U, so U times V is N for every N and every U that is not zero. Algebra will not let the second step disagree with the first. The multiplication confirms the arithmetic and nothing else about the scheme.

A real check needs two sides drawn from two places, and there is one sitting in plain view. Add the 3,80,000 folio lines together and the total ought to be 1,20,00,00,000 units. The total of the folio lines comes off the register. The units figure used in the division above comes from the scheme's books. Two sources, so the two can differ, and if they ever do, somebody has to find out why before anything is published. The matches of that kind, and how many of them a scheme runs, are covered under reconciliation; what matters here is only the shape.

One small division shows how tight such a check is, and it is worth doing by hand. Dividing 1,20,00,00,000 units by 3,80,000 folios gives 60000 over 19. The division runs to 3,157.894736 and keeps going, non-terminating with 19 in the denominator. A register of 3,80,000 equal rows is therefore not merely unlikely, it is arithmetically impossible. Test it in both directions and watch it break each way: 3,80,000 rows of 3,157.894 units come to 1,19,99,99,720, which is 280 units short of the total, and 3,80,000 rows of 3,157.895 units come to 1,20,00,00,100, which is 100 units over. A check that fails in both directions is a check worth running.

What is each record able to establish on its own?

A great deal, inside its own boundary. The custody record can show that the scheme has in its charge what it says it has: so many of security P, so many of security Q, this much cash, this much due in and this much due out. Folio A carries one figure and no other, so the unit register can show that a holder is standing on the balance the holder thinks they are standing on. Both are strong statements, both can be tested, and a scheme unable to make either would be in serious trouble.

Now put the two side by side and see what has gone missing. Neither one, by itself or in company with the other, can show that the value per unit is right. A register right to the last unit and a custody record right to the last security sit perfectly comfortably beside a wrong number in between. Valuation is not something a register knows about. The custody record knows how many of security P have been bought and not what one of them is worth. Worth arrives from a valuation method written down beforehand and applied by somebody else, on which neither of the two records has any view.

Exactness about what this arrangement achieves calls for equal exactness about what it does not. Splitting the two roles is a control over who is able to move what; it is not a check on what the holdings go on to do. The value per unit can fall a long way with both records kept immaculately, and no amount of splitting the task takes a paise off that. The split governs who may move a security and who may move a unit balance. About what a scheme achieves the split says exactly nothing, and the two ideas are never to be run together.

Why can a record never check itself?

Because a record hangs together by the way it was built. Total the columns of a register a second time and they tie. They tied on the day they were written, and nothing outside has been asked about anything since. Finding a record consistent with itself shows only that it was assembled consistently. Consistency is a fact about records as a species, not evidence about the world. A shopkeeper who adds his day book up twice has confirmed his addition. He has not confirmed that anybody handed him money.

None of this is an accusation against either party, though it can read like one. A meticulous keeper and a slapdash keeper both turn out records that hang together, and that is precisely the difficulty: an internal check cannot distinguish the two. The property sits with records, not with people. So the checking in a scheme is laid out across parties instead of within any single one, and the useful question about a check is never how thorough it was, but where its second side came from.

Which leads straight to the question the whole comparison has been circling.

What would happen if one party did all three jobs?

One such party would turn out a beautiful set of records, and that is the trouble. Put one party in charge of the securities, of saying what they come to and of the roll of who they answer to, and the three records it produces will agree with each other flawlessly. A single hand wrote all three and had every reason to make them tie. Flawless agreement is the cheapest thing such an arrangement can manufacture, and nobody standing outside it could test the result at all. There is no second source anywhere to set against the first.

Notice that this argument needs no bad actor in it. The separation of rolesSplitting jobs between parties so that no one of them can complete a task end to end alone. is not built on a suspicion that any particular keeper would behave badly. A scrupulous single keeper and a careless one hand over the same internally consistent bundle, and an outsider has no way of telling the two apart. Take the household version. A cousin who does the shopping, keeps the accounts and holds the cash box will produce accounts that balance every single month. The accounts balance whether or not anything is amiss, so nobody learns a thing from the fact that they balance.

Holders do not rely on a promise from any party. No single party is ever in a position to write all three records, so agreement between them has to be earned across a boundary rather than arranged inside one set of books. That is a control over who is able to move what. The control says nothing at all about what the holdings do.

How do the two ends tie on the Girnar Large Cap Equity Fund?

Both ends stated, then the join performed, all in a single run. The custody end comes first. The securities the scheme has bought are priced by the method written down beforehand, the cash lodged at a bank is added, everything due in is added, everything due out is taken away, and the result is Rs 42,00,00,00,000/-. The same figure is Rs 4,200 crore of net assets set out in full rupees. Then the register end. The 3,80,000 folio lines add to 1,20,00,00,000 units, or 120.00 crore written the same way.

Now the join, already met once above. Rs 42,00,00,00,000/- over 1,20,00,00,000 units is Rs 35.00 a unit exactly, and 1,20,00,00,000 units at Rs 35.00 hands back Rs 42,00,00,00,000/-. One multiplication is the whole of what ties the two records together, and anybody able to do it can say, to the paise, what each end of a scheme is asserting.

The last two rows of the table repay slow reading. Suppose every security in the custody record is right, every unit in the register is right, and the pricing is out by Rs 1.20 crore, that is Rs 1,20,00,000/-. Both records stay exactly as right as they were. Nothing in either has moved. Spread across 1,20,00,00,000 units, the error comes to one paise a unit with nothing left over, so what goes out is Rs 35.01 where Rs 35.00 belonged. A whole paise of published value has shifted without one entry in either record being wrong.

Brought down to a holder it stops being abstract. An average folio on this scheme carries 60000 over 19 units, or 3,157.894736 and running, so a paise a unit works out at 60000 over 19 paise, about Rs 31.58 on that average folio. A small arithmetic curiosity is sitting there: since the error is one paise a unit exactly, the amount in paise is numerically the same fraction as the unit count, 60000 over 19 read either way. The same division also produces the average folio value of about Rs 1,10,526/-, being 3,157.894736 units at Rs 35.00.

Which leaves the closing reading of the whole comparison, and it is stricter than most readers expect. Between them the two records fix what is there and whose claim it answers, and neither, on its own or in company, fixes what any of it comes to. Hence a pricing method written down beforehand instead of settled in the moment, put to work by a party keeping neither record and then tested by a fourth. The two roles differ by the work each one does, never by price and never by size.

Which three questions describe any pooled arrangement?

Peeled off this scheme, the shape travels anywhere. Any arrangement in which money is pooled, be it a scheme, a chit, a housing society sinking pot or the crockery room above, is described by the answers to three questions. Who keeps the thing. Who keeps the record of whose claim it answers. Who decides what it is worth. The three answers mark out where the boundaries run, and therefore which checks are capable of coming out wrong. No description of the arrangement says more.

Using the three answers is the useful part, and the use is not an alarm. One party answering two of the three does not mean something has gone wrong, and it certainly does not mean anybody is misbehaving. The overlap means a check that used to cross a boundary now sits inside one set of books, so the check has lost its ability to surprise. A structural fact of that kind, held in mind, makes the follow up question a sensible one rather than a suspicious one.

Who sets the registration and record keeping requirements?

The market regulator. Registration for either role, what has to be kept, what has to be reported and every condition hanging off any of it is fixed by the Securities and Exchange Board of India (SEBI) and published at sebi.gov.in, and requirements of that kind get revised. A holding kept in dematerialised form is described by the two depositoriesA body through which securities and units can sit in electronic form instead of on paper.: the National Securities Depository Limited (NSDL), at nsdl.co.in, with Central Depository Services (India) Limited (CDSL) doing the same at cdslindia.com.

No registration requirement disturbs the split of work between the two record keepers. Whatever today's registration condition happens to say, securities still sit in one party's charge and the roll of folios and unit balances still sits with another, because that split is the point of the arrangement rather than a by product of some particular rule. Requirements change how a party qualifies and what it must keep on file. No requirement changes which record answers which question.

So which of the two registers is the senior one?

Neither, and the question is more interesting than the answer. A reader who arrives wanting to be told which of the two matters more has asked something the arrangement is built to have no answer to. Ranking them misses the point twice over. The two records answer different questions, and a ranking between different questions is meaningless. Any argument for making one of them senior is also an argument for eventually letting it absorb the other, and absorption is the outcome the division exists to prevent.

The comparison settles four things, one line each. One party keeps a thing and the record of that thing, the other keeps the record and nothing else, neither can produce the other's answer because neither holds the other's input, and the single number where they meet is the one thing neither of them can vouch for. Those four statements are enough to walk into any pooled arrangement, ask the three questions, and know what it is inside a minute.

Girnar Asset Management Limited, the Girnar Large Cap Equity Fund, Kalyani Bhagat and Sohail Merchant are invented.
Educational material. Not advice on any investment, tax, budget or market position.